A Comparative Analysis of Machine Learning Models for Intrusion Detection in IoT Networks
Bildiriler > Tam Metin Bildiri · 2025
Yazarlar: Merve Candan, Özgü CAN
5th International Conference on Electrical, Computer and Energy Technologies (ICECET), Fransa, Paris, Uluslararası, Sözlü Sunum, Tam Metin Bildiri, 2025-2025, ss 877-882
Özet
<p>The widespread adoption of Internet of Things</p><p>(IoT) devices across various domains has significantly enhanced</p><p>connectivity and automation. However, the inherent security</p><p>vulnerabilities of these devices make them prime targets for</p><p>cyberattacks. Ensuring the security of IoT ecosystems necessitates</p><p>robust mechanisms for the detection and prevention of malicious</p><p>activities. Intrusion Detection Systems (IDS) play a crucial role</p><p>in this regard by ceaselessly tracking network traffic t o identify</p><p>anomalous behavior and potential threats. However, conventional</p><p>IDS approaches often struggle to detect sophisticated and</p><p>evolving attack patterns in IoT environments. Machine learning</p><p>(ML) techniques have emerged as a promising solution for</p><p>enhancing IDS capabilities by enabling adaptive and data-driven</p><p>threat detection. This study conducts a comprehensive evaluation</p><p>of various ML algorithms for IDS using the RT-IoT2022 dataset.</p><p>Specifically, t he p erformance o f R andom F orest, D ecision Tree,</p><p>K-Nearest Neighbors (KNN), Logistic Regression, Naive Bayes,</p><p>and Gradient Boosting classifiers i s a ssessed. T hese models</p><p>are trained and tested, and their classification p erformance is</p><p>evaluated using accuracy, precision, recall, and F1-score metrics.</p><p>The experimental findings i ndicate t hat t he G radient Boosting</p><p>algorithm outperforms other models, achieving the highest</p><p>accuracy, precision, recall, and F1-score. While Random Forest</p><p>and Decision Tree models also demonstrate strong classification</p><p>performance, they exhibit higher computational costs in terms</p><p>of processing time. The results of this study provide valuable</p><p>insights for cybersecurity practitioners in selecting suitable</p><p>ML-based IDS solutions for securing IoT networks. Furthermore,</p><p>these findings s erve a s a f oundation f or f uture r esearch aimed</p><p>at optimizing IDS frameworks to enhance the resilience of IoT</p><p>ecosystems against emerging cyber threats.</p>